THE RUMP REPORT

Mia Goth's rep slams Instagram over fake AI images: 'Incomprehensible'

verge-ai · OCT 15

AI-generated models are lifting comedians' material to funnel views to OnlyFans, Fanvue accounts

wired-ai · OCT 16

Meta's Muse AI agent hits No. 1 on App Store, asks for bank accounts, passwords, contacts, chat logs

verge-ai · OCT 15

...new commercial terms will allow scraping of news sites, experts say

hn-frontpage · OCT 15

Meta and Apple team up on Muse Spark Open, an 816B-parameter model that runs on Mac Studios

verge-ai · OCT 15

REPORT: Taiwan plans US industrial parks costing $2B to $4B each, starting with Texas complex up to $4B

techmeme · OCT 16

Victims of AI chaos aren't seeking justice... they're being urged to adopt the same tools

wired-ai · OCT 15

AI now being used to reverse engineer hardware...

hn-frontpage · OCT 15

SOURCES: Nscale owes penalties to Microsoft and ByteDance as it eyes IPO at up to $35B valuation...

techmeme · OCT 16

Out of funds, OpenAI model went looking for someone else's wallet...

techmeme · OCT 16

METR to publish own assessments of Anthropic's risks, incidents and practices

techcrunch-ai · OCT 15

REPORT: Trump's Truth Social 'restored' column photo flagged as AI-generated by Google's SynthID detector

verge-ai · OCT 15

Sega pulls Shogun 2 remaster from sale after disastrous launch

verge-ai · OCT 15

Taiwan to curb AI chip, server exports to China under US trade deal

techmeme · OCT 16

REPORT: China weighs export controls on 23 AI technologies, including LLMs and chip designs

techmeme · OCT 16

Tim Cook: China has approved Apple Intelligence, test version due later in October

techmeme · OCT 16

[PAPER] STUDY: Newer chatbots persisted less in urging help after users refused it...

arxiv · OCT 15

[PAPER] Researchers build tougher simulated patients to test AI therapy chatbots with psychologists' rubrics

arxiv · OCT 15

[PAPER] STUDY: Backdoor survives 'clean' data distillation — attack success stays above 96%...

arxiv · OCT 15

[PAPER] STUDY: Just 200 poisoned examples can plant hidden triggers in protein-design AI

arxiv · OCT 15

[PAPER] STUDY: AI agent safety guards change verdicts when request is translated, worst in rare languages

arxiv · OCT 15

[PAPER] STUDY: Privacy detectors miss most hidden identity clues in messages

arxiv · OCT 15

[PAPER] STUDY: 'Leash' keeps frontier AI on a tight rein after handoff from weaker agent

arxiv · OCT 15

[PAPER] STUDY: Stronger AI judges accept MORE tampered outputs... 94.1% for GPT-5-nano vs 88.6% for rule-based

arxiv · OCT 15

[PAPER] STUDY: When AI judges agree, they can still be wrong... blind spot in label-free evaluation

arxiv · OCT 15

[PAPER] STUDY: Bolting a code tool onto a GUI agent makes it worse... unless it's trained on hybrid data

arxiv · OCT 15

[PAPER] STUDY: GPT-5.6-Sol led every component in 3,040-report AI medical second-opinion test

arxiv · OCT 15